The architecture of a fault-tolerant cached RAID controller
ISCA '93 Proceedings of the 20th annual international symposium on computer architecture
Performance of RAID5 disk arrays with read and write caching
Distributed and Parallel Databases - Special issue on disk arrays
Architectures and algorithms for on-line failure recovery in redundant disk arrays
Distributed and Parallel Databases - Special issue on disk arrays
Approximate solutions for M/G/1 fork/join synchronization
WSC '94 Proceedings of the 26th conference on Winter simulation
Performance Analysis of RAID5 Disk Arrays with a Vacationing Server Model for Rebuild Mode Operation
Proceedings of the Tenth International Conference on Data Engineering
Surveyors' forum—high-performance secondary memory
ACM Computing Surveys (CSUR)
RAID5 Performance with Distributed Sparing
IEEE Transactions on Parallel and Distributed Systems
Systems Support for Preemptive Disk Scheduling
IEEE Transactions on Computers
Hi-index | 0.00 |
We consider a cached RAID5 disk array which partially consists of non-volatile storage (NVS) providing a fast-write capability. Blocks written in this manner are destaged from the NVS cache onto disk asynchronously and at a lower priority than disk read requests. We consider three queueing disciplines to prioritize read misses with respect to destages: (i) Non-preemptive priorities. (ii) Preemptions only after completion of a seek (split-seek option). (iii) Preemption during the latency phase (split-latency option). It can be concluded from the simulation results for two workloads that unlike the FCFS policy, the increase in read response time for all three disciplines is quite moderate up to the point of full disk utilization. This improvement in read response time is at the cost of some wasted processing, but it is shown by applying the vacationing server model to the split-seek option that the reduction in the maximum throughput with respect to the FCFS policy is negligibly small.